def create_prover(options: deephol_pb2.ProverOptions) -> Prover:
"""Creates a Prover object, initializing all dependencies."""
theorem_database = io_util.load_theorem_database_from_file(
str(options.path_theorem_database))
tactics = io_util.load_tactics_from_file(str(options.path_tactics),
str(options.path_tactics_replace))
if options.action_generator_options.asm_meson_no_params_only:
tf.logging.warn(
'Note: Using Meson action generator with no parameters.')
action_gen = action_generator.MesonActionGenerator()
else:
predictor = get_predictor(options)
emb_store = None
if options.HasField('theorem_embeddings'):
emb_store = embedding_store.TheoremEmbeddingStore(predictor)
emb_store.read_embeddings(str(options.theorem_embeddings))
assert emb_store.thm_embeddings.shape[0] == len(
theorem_database.theorems)
action_gen = action_generator.ActionGenerator(
theorem_database, tactics, predictor,
options.action_generator_options, options.model_architecture,
emb_store)
hol_wrapper = setup_prover(theorem_database)
tf.logging.info('DeepHOL dependencies initialization complete.')
if options.prover == 'bfs':
return BFSProver(options, hol_wrapper, action_gen, theorem_database)
return NoBacktrackProver(options, hol_wrapper, action_gen,
theorem_database)
def test_action_generator_theorem_list_parameter_tactic(
self, use_embedding_store):
"""Checks max_theorem_parameters parameters are passed for a thmlist tactic.
Args:
use_embedding_store: True if the embedding store should be used.
"""
max_parameters = self.options.max_theorem_parameters
emb_store = None
thmlist_param_tactic = deephol_pb2.Tactic(
name='TAC', parameter_types=[deephol_pb2.Tactic.THEOREM_LIST])
dummy_theorem = proof_assistant_pb2.Theorem(name='THM',
conclusion='foo')
theorem_database = proof_assistant_pb2.TheoremDatabase()
theorem_database.theorems.extend([
proof_assistant_pb2.Theorem(name='THM%d' % i, conclusion='foo')
for i in range(2 * max_parameters + 1)
])
if use_embedding_store:
emb_store = embedding_store.TheoremEmbeddingStore(self.predictor)
emb_store.compute_embeddings_for_thms_from_db(theorem_database)
action_gen = action_generator.ActionGenerator(
theorem_database, [thmlist_param_tactic], self.predictor,
self.options, self.model_architecture, emb_store)
test_theorem = theorem_database.theorems[2 * max_parameters]
actions_scores = action_gen.step(
self.node, prover_util.make_premise_set(test_theorem, 'default'))
self.assertStartsWith(actions_scores[-1].string, 'TAC')
self.assertEqual(max_parameters,
actions_scores[-1].string.count('THM'))
def test_action_generator_unknown_parameter_tactic(self):
unknown_param_tactic = deephol_pb2.Tactic(
name='TAC', parameter_types=[deephol_pb2.Tactic.UNKNOWN])
action_gen = action_generator.ActionGenerator(self.theorem_database,
[unknown_param_tactic],
self.predictor,
self.options,
self.model_architecture)
actions_scores = action_gen.step(self.node, self.test_premise_set)
self.assertEqual(0, len(actions_scores))
def test_action_generator_no_parameter_tactic(self):
no_param_tactic = deephol_pb2.Tactic(name='TAC')
action_gen = action_generator.ActionGenerator(self.theorem_database,
[no_param_tactic],
self.predictor,
self.options,
self.model_architecture)
actions_scores = action_gen.step(self.node, self.test_premise_set)
self.assertEqual(1, len(actions_scores))
self.assertEqual(actions_scores[0].string, 'TAC')
def test_action_generator_theorem_parameter_tactic(self):
thm_param_tactic = deephol_pb2.Tactic(
name='TAC', parameter_types=[deephol_pb2.Tactic.THEOREM])
action_gen = action_generator.ActionGenerator(self.theorem_database,
[thm_param_tactic],
self.predictor,
self.options,
self.model_architecture)
actions_scores = action_gen.step(self.node, self.test_premise_set)
self.assertEqual(1, len(actions_scores))
expected = 'TAC ' + theorem_fingerprint.ToTacticArgument(
self.theorem_database.theorems[0])
self.assertEqual(expected, actions_scores[0].string)
def test_action_generator_hol_light_tactics_sanity_check(self):
"""HolLight tactics sanity test.
This is a sanity check to ensure action generator works with actual HolLight
tactics on which the test model was trained.
"""
hollight_tactics = load_tactics(HOLLIGHT_TACTICS_TEXTPB_PATH)
action_gen = action_generator.ActionGenerator(self.theorem_database,
hollight_tactics,
self.predictor,
self.options,
self.model_architecture)
actions_with_scores = action_gen.step(self.node, self.test_premise_set)
for action, score in sorted(actions_with_scores,
key=lambda x: x.score):
tf.logging.info(str(score) + ': ' + str(action))
self.assertIn('EQ_TAC', [action for action, _ in actions_with_scores])
